1: I've been trying to read quite a lot about ISO settings, and so far as I can work out: you don't gain any more at all in sensitivity beyond about ISO800-1600, possibly even less; but you lose 50% in dynamic range per ISO doubling; and background noise covers a bigger fraction of that dynamic range. I'm still not perfectly sure about the details, but you may be better off shooting at ISO 800/1600 rather than 1600/3200 and then stretching with 'curves' in Photoshop? Perhaps that would make it a little less noisy?
This is an interesting comment and one I'd like to see explored and clarified a bit more because it differs from my own understanding, which may well be wrong.
My understanding is that dynamic range is basically the range that constitutes the "intelligible" part of the signal and by increasing the ISO, a large part of the dynamic range loss actually comes from the increased noise drowning out the signal making it "unintelligible". The remainder would be highlights that get lost due to clipping that wouldn't otherwise take place. So unless I'm mistaken, you don't lose out with decreased dynamic range and increased noise, I think they're one and the same.
The flip side to that is that some cameras, (some?) Nikons at least, clip values within a certain range of 0, so by upping the ISO and pushing the signal outside of that range you'd actually be increasing your dynamic range.
I saw a really good article on this a while ago that explained it much better than I can, but I have no idea where it is now. It indicated that there's a certain sweet spot with dSLRs where you get the most dynamic range... ISO 400 was a good spot for one of the particular cameras.
But does any of this, dynamic range etc take stacking into account? If as Ray is doing, takes many 8 bit jpegs and stacks them in DSS, the final image is 32 bit anyway.
With the testing I did on my Pentax K-x, I couldn't see any difference between 1 x 4 minute sub at ISO400, 2 x 2 minute subs stacked at ISO800 or 4 x 1 minute subs stacked at ISO1600. I never took the test over ISO 1600 but assume (perhaps incorrectly) that 8 x 30 seconds stacked at ISO3200 or 16 x 15 seconds at ISO 6400 would be much the same.
Now I have read that longer subs are better, but not sure if that applies to DSLR's that are not cooled. Long subs on DSLR's suffer from thermal noise.
Perhaps without realising it, Ray processes in 32 bit by doing the processing in house with DSS. That's an advantage to processing in 16 bit say with photoshop or worse in 8 bit with other software.
Two things come to mind that would help with this image though, more subs (the more the merrier lol) and taking flats to reduce the vignetting.
Thanks for your comments Andy and Lee. From what I have read, the ISO
sweet spot for most Canon DSLRs seems to be between 800 and 1000.
I am lowering my ISO as my tracking gets better. Now mostly using 1600. I know nothing about losing dynamic range at higher ISOs, but I
do know that images are much brighter, and more stars are visible, with
each ISO step up to 6400. {my camera's max} Beyond that figure I
wouldn't know. Obviously, the noise increases too.
I cannot see the haloes you mention, even at 400x{ when the pic is pixelated anyway}, but my eyes are not what they were. Assuming that they are there, maybe I've oversharpened a little. I'll have a look in DPP
at the unsharp mask setting. I use the ancient PS 7, so don't have the
features you mention. I don't use raw; too much stuffing around, and
humungous files. I don't use separate darks, flats, etc. I'm just
trying to produce the best results I can using basic methods.
raymo
Thanks for your contribution Kevin; you're spot on with the more subs comment, and I can say that I have taken two 30sec 6400 subs, and
four 30 sec ones, and couldn't see the slightest difference; even the noise looked pretty much the same. I can see differences in star densities in
the four corners of the image, but can't see any vignetting. It would
surprise me if I could, because the image is heavily cropped. Perhaps my
eyes are even worse than I thought. Incidentally, my 80mm achro
suffers badly with vignetting.
raymo
I find that iso 3200 is the best compromise for my 1100d canon, but it has worked really well, with low noise at 6400 on a cold night. Vignetting is a camera field of veiw/ sensor size thing. My 8" f6 newt had bad vignetting with a 35mm film camera at prime focus but is good with the canon 1100d crop sensor
Interesting points everyone! I certainly don't wish to disparage your eyes raymo , so perhaps best I give examples. The first image is a blow-up crop of part of your image, showing the darkened 'haloes' round the moderately bright stars (not on the brightest diffraction-spiked stars). You may need to magnify the image to see the haloes clearly. I get the same thing out my camera, I'm >90% sure that's due to unsharp masking put on images by a setting in the camera, and leads to hard edges round some stars, and dark haloes round others. The second image is a crop from one of my image subs, un-processed, but with the default 'unsharp mask' set by Digital Photo Professional to 3. This is how the jpegs look. The 3rd image is identical, but for the unsharp mask set to zero. The haloes disappear without the strong unsharp mask.
Just this evening, I discovered how to have your camera set the unsharp masking to zero by default (so you don't have to process it out). In the menu on the back of the camera, set "picture style" from 'standard' to 'neutral'. If that works for you, it would save you doing any fancy processing!
I really need to look out the detailed discussions of ISO and dynamic range - I don't really want to be making unsupported comments!! I'll look for the articles I was reading, beacause they were excellently supported by data. The technical part if I recall was that with the ISO at 400-800, each photon filled about 1 unit of the well in the detector. Until I can find the right article, this link has some exposure-ISO data - see discussions and graphs in the first answer. I *think* that it is read noise that limits lower ISO (and the need for longer exposures of course), and then dynamic range limits the higher ISOs, and the noise becomes a larger fraction of the dynamic range. Stacking images increases the dynamic range of your whole image, but only if the stars/pixels aren't already saturated.
I'm reasonably sure that more exposure in each sub = more signal, better signal-to-noise and generally better images, but again I need to find the appropriate data to prove this - I can recall an article with a series of California Nebula images demonstrating this, but can't find it just now. Obviously that' limited by how well your set-up tracks the stars!
Sorry for the long comment and hi-jacking of your lovely image, maybe we need to start a separate thread if discussion continues?
Thanks again Andy, I've set the picture style to neutral, and will see what happens when the weather finally lets me do some imaging. I'm
inclined to think that maybe I'm slightly overdoing the unsharp masking
myself. I saw the haloes, but for some reason they are much fainter
on my laptop than in your images.
Sorry Rick, some of it is over my head too.
raymo
Last edited by raymo; 11-07-2014 at 11:25 PM.
Reason: more info
hey chaps, here's a very informative site re. exposure vs no. of subs. It shows that 20 is a good number, massive law of diminishing returns after 20. Clearly, exposure length is king.
I still think there is a place for many subs. As in the hundreds if not thousands. What I've found is that for every double number of subs, there is half the noise. For instance 2 subs stacked is half the noise of a single one, 4 is half the noise of 2, 8 is half the noise of 4 etc right up to 1000 is half the noise of 500! So apart from killing the poor camera, why stop? lol.
Less noise means the image can be stretched more in the shadow areas so brings out fainter detail. (Taking the longest subs one can reasonably do without trailing of course)
Sorry Ray, I'm afraid we've totally hijacked your thread now.
No worries Kevin, it's still interesting. There is a member [who's name I
don't remember] who occasionally posts stunning pics here, that are the result of 250 or more subs.
raymo
I went and dug out the links I was looking at (on another machine), and found that one was Simmo's first one!
Another one, by Craig Stark, has really good data about ISO and dynamic range in an article entitled "Profiling the long exposure performance of a Canon DSLR" about halfway down the article, (it's pretty technical in other parts too). Look for the "Dynamic range vs ISO" section. The take-home message was a loss of 50% dynamic range per ISO stop, and no increased actual sensitivity somewhere beyond approximately ISO 400 (the exact value will vary with model).
Thanks again for the information, which I will read; and for the compliment. I don't know if I have the meaning of sensitivity wrong,
but if I make exposures of the same length using 400-800-1600-3200,
and 6400 ISO, the images get progressively brighter, and show more and more stars and detail. I assume that this would continue if my camera's ISO went higher. I realise that as the ISO rises, so does the noise, but it still seems to me that sensitivity increases with the ISO setting.
raymo
, so perhaps best I give examples. The first image is a blow-up crop of part of your image, showing the darkened 'haloes' round the moderately bright stars (not on the brightest diffraction-spiked stars). You may need to magnify the image to see the haloes clearly. I get the same thing out my camera, I'm >90% sure that's due to unsharp masking put on images by a setting in the camera, and leads to hard edges round some stars, and dark haloes round others. The second image is a crop from one of my image subs, un-processed, but with the default 'unsharp mask' set by Digital Photo Professional to 3. This is how the jpegs look. The 3rd image is identical, but for the unsharp mask set to zero. The haloes disappear without the strong unsharp mask.
